[PATCH] D58729: Emit boxed expression as a compile-time constant if the expression inside the parentheses is a string literal

[John McCall via Phabricator via cfe-commits]

rjmccall added inline comments.


================
Comment at: test/SemaObjC/boxing-illegal.m:70
+  s = @(L"abc"); // expected-error {{illegal type 'int *' used in a boxed expression}}
+  s = @("\pabc"); // expected-error {{illegal type 'unsigned char *' used in a boxed expression}}
+}
----------------
ahatanak wrote:
> ahatanak wrote:
> > rjmccall wrote:
> > > I don't know what `\p` is supposed to be or why it apparently changes the type of the literal to `unsigned char *`, but none of these are ordinary string literals that are invalid as UTF-8.  I mean something like "\xFF", which still has type `char *` but will fail to parse as UTF-8, which will cause normal boxing to fail and return `nil`.
> > Fixed and added test cases. Boxed expressions now have to be valid UTF-8 string literals in order to be emitted as compile-time constants.
> > 
> > If the string literal in a boxed expression is an invalid UTF-8 string, should we reject it the same way we reject other kinds of string literals (e.g., UTF-16)?
> Or at least warn about it.
Thanks, fix looks good.  I think a warning would be extremely reasonable. 


Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D58729/new/

https://reviews.llvm.org/D58729